The command actually trains the whole pipeline including mining a
transliteration corpus from the parallel corpus. The language model for the
transliteration module is trained on characters (target-side of the mined
transliteration corpus). But as Anoop mentioned, if you have a
transliteration corp
Hi,
Did the miner create a transliteration corpus? I see you are using corpus
with .fr extension. This may cause transliteration miner to not run.
Transliteration is only enabled for languages where writing scripts are
different.
Can you send *.STDERR/STDOUT files and training corpus with alignme
Hi Rofael,
The script runs transliteration mining over parallel corpus, for which it
requires word alignments. Please see the link below for details
http://www.statmt.org/moses/?n=Advanced.OOVs#ntoc4
Nadir
On Mon, May 1, 2017 at 7:00 PM, wrote:
> Send Moses-support mailing list submissions t
Hi WULongski,
When you enable
[TRAINING]
transliteration-module = "yes"
in the EMS, it simply trains transliteration model from your word-aligned
parallel corpus. This includes i) mining transliteration corpus, ii) then
training the entire phrase-based pipeline over character corpus that was
apply, please send an email to care...@qcri.org and cc svo...@qf.org.qa
For further inquiries, please contact Hassan Sajjad, Research Scientist at
QCRI (hsaj...@qf.org.qa).
Best,
Nadir Durrani
___
Moses-support mailing list
Moses-support@mit.edu
http
Hi Sanjna,
You are getting the error in threshold.pl
This means that Miner did not function correctly and the *.probs file which
the threshold script takes is empty. Are you running the training manually
or through train-transliteration-module.pl? Please make sure to run the
cleaning script on yo
Hi Lane,
I hope upgrading to a newer version will have solved your problem. The
training script
OSM-Train.perl
takes both
*'srilm-dir=s' => \$SRILM_DIR,'lmplz=s' => \$LMPLZ,*and if the user doesn't
specify any, which should trigger with
*operation-sequence-model-settings = ""*
then
if (!
Hi Ergun,
processPhraseTable is no longer supported by Moses. But I see that
Phil Williams has already fixed this problem in transliteration
module, by changing
`$MOSES_SRC/scripts/training/filter-model-given-input.pl
$TRANSLIT_MODEL/evaluation/$eval_file.filtered
$TRANSLIT_MODEL/evaluation/$eva
Hi Emna,
Moses can mine transliteration pairs from a word-aligned parallel
data, and build transliteration model from it. You can then use the
transliteration model to translate OOV words either in the
post-decoding step or on the fly.
Here's some information on how to enable transliteration modu
Hi Miguel,
If you just want to use language model over word-classes, here's a set
of instructions to do that
http://www.statmt.org/moses/?n=Advanced.Models#ntoc4
Cheers,
Nadir
On Tue, Mar 31, 2015 at 7:25 PM, wrote:
> Send Moses-support mailing list submissions to
> moses-support@mit.
Hi,
We have recently added a transliteration module in Moses. Here's how to use it:
http://www.statmt.org/moses/?n=Moses.AdvancedFeatures#ntoc5
When enabled the model will try to mine transliteration pairs from the
word-aligned parallel corpus and build a transliteration model over
the resulting
Hi,
There is going to be a transliteration module in Moses soon. We (me
and Hassan Sajjad) are working on it.
Cheers,
Nadir
On Thu, Nov 28, 2013 at 10:17 AM, wrote:
> Send Moses-support mailing list submissions to
> moses-support@mit.edu
>
> To subscribe or unsubscribe via the World Wi
The recent version of OSM-decoder from LMU-Munich uses discontinuous
source-side phrases. We used it in this year's WMT campaign. Details
on phrase extraction can be looked at in
http://www.statmt.org/wmt13/pdf/WMT13.pdf
It gives improvements although not consistently which I suppose is
also true
h the boost 1.53 libraries it finds in
> > /my/custom/boost.
> >
> > I'd like to modify the bjam files so that if --with-boost is provided, it
> > will add a -I flag to g++ so that the boost headers in the provided path
> > get searched before the system include paths. Unfortu
14 matches
Mail list logo
